caveman-review

To generate concise code review comments, you provide a PR diff to be analyzed, resulting in ultra-brief, actionable one-line comments for each finding. This approach reduces noise while preserving the exact line references and specific fixes needed.

Automated PR review comments use a consistent format of L<line>: <problem>. <fix>, or <file>:L<line>: ... for multi-file diffs. This enforces exact problem descriptions and concrete fixes, supporting optional severity prefixes like bug, risk, nit, or q.

Yes, you can review multi-file diffs without losing exact line references. The review process preserves the original context and supports the <file>:L<line> format, ensuring deterministic outputs with accurate line tracking across all changed files.

You can add severity levels to automated linting comments by applying optional severity prefixes such as bug, risk, nit, or q. This categorizes the one-line comments generated for each finding in the PR diff, highlighting the urgency or type of issue.

The best way to reduce noise from PR code review feedback is to use ultra-compressed, one-line comments that enforce a strict problem-and-fix format. This ensures every comment is actionable and deterministic, eliminating unnecessary explanations.

Yes, the review automation auto-triggers when reviewing pull requests. It streamlines the process by automatically analyzing the provided diff and generating terse, actionable one-line comments for each finding without manual activation.